摘要
针对目前基于核密度估计的去噪算法在核函数参数的选取上未能充分体现散乱点云数据的表面特征,提出一种改进的去噪算法。以当前点法向量与其邻域内点的法向量构造的差向量作为核函数的参数,引入面积权重进行光顺,通过构造空间单元格的最大连通域剔除离群点,结合K-近邻搜索建立点云之间的拓扑关系,以改进的高斯函数作为核函数计算当前点的影响值。实验结果表明,该算法在有效去除表面噪声和离群点的同时,能够较好保留模型的细节特征。
An improved de-noising algorithm was proposed considering that the current algorithms based on kernel density esti-mation fail to fully reflect the scattered point cloud on the selection of kernel function parameters.Difference vectors constructed with normal vectors of the point and that of its neighborhood were used as the kernel function parameters while introducing area weight to smooth.Maximally connected domain of the space grid was obtained to eliminate outliers,and K-nearest neighborhood search was combined to establish the topological relations,and the improved Gaussian function was taken as kernel function to calculate influence of the point.The experimental results indicate that the algorithm can effectively remove surface noise and out-liers and preserve detail features of model.
出处
《计算机工程与设计》
北大核心
2015年第5期1285-1289,共5页
Computer Engineering and Design
基金
四川省教育厅基金项目(13ZB0184)
核废物与环境安全国防重点学科实验室开放基金项目(13ZXNK07)
关键词
点云去噪
高斯核函数
空间单元格
K-近邻
协方差分析
point cloud de-noising
Gaussian kernel function
space grid
KNN
analysis of covariance
